2. I.INTRODUCTION
OPTIMIZATION
It is defined as follows: choosing the best element from some set of available
alternatives.
• In Pharmacy word “optimization” is found in the literature referring to any study of
formula.
• In development projects pharmacist generally experiments by a series of logical
steps, carefully controlling the variables and changing one at a time until
satisfactory results are obtained. This is how the optimization done in
pharmaceutical industry.
2
3. II. OPTIMIZATION PARAMETERS
There are two optimization parameters
1.Problem Types
2.Variables
• PROBLEM TYPES -There are two general types of optimization problems:
1. Unconstrained
2. Constrained
In unconstrained optimization problems there are no restrictions. For a given pharmaceutical
system one might wish to make the hardest tablet possible. This making of the hardest tablet is the
unconstrained optimization problem. The constrained problem involved in it is to make the hardest
tablet possible, but it must disintegrate in less than 15 minutes.
3
4. • VARIABLES - The development procedure of the pharmaceutical formulation involves
several variables. Mathematically these variables are divided into two groups.
1.Independent variables
2.Dependent variables
The independent variables are under the control of the formulator. These might include
the compression force or the die cavity filling or the mixing time. The dependent
variables are the responses or the characteristics that are developed due to the
independent variables. The more the variables that are present in the system the more
the complications that are involved in the optimization.
4
5. Once the relationship
between the variable and
the response is known, it
gives the response surface
as represented in the Fig. 1.
Surface is to be evaluated to
get the independent
variables, X1 and X2, which
gave the response, Y. Any
number of variables can be
considered, it is impossible
to represent graphically, but
mathematically it can be
evaluated.
5
6. III. CLASSICAL OPTIMIZATION
•Classical optimization is done by using the calculus to basic problem to find the
maximum and the minimum of a function.
•The curve in the Fig. 2. represents the relationship between the response Y and the
single independent variable X and we can obtain the maximum and the minimum. By
using the calculus the graphical represented can be avoided. If the relationship, the
equation for Y as a function of X, is available [Eq. (1)]:
Y = f(X)
Figure 2. Graphic location of optimum (maximum or minimum)
6
7. • When the relationship for the response Y is given as the function of two independent
variables, X1 and X2 ,
Y = f(X1, X2)
•Graphically, there are contour plots (Fig. 3.) on which the axes represents the two
independent variables, X1 and X2, and contours represents the response Y.
Figure 3. Contour plot. Contour represents values of the dependent
variable Y
7
8. V. APPLIED OPTIMIZATION METHODS
There are several methods used for optimization. They are
Evolutionary
Operations
The Simplex
Method
The Lagrangian
Method
Search Method
Canonical
Analysis
8
9. EVOLUTIONARY OPERATIONS
• One of the most widely used methods of experimental optimization in fields
other than pharmaceutical technology is the evolutionary operation (EVOP).
• This technique is especially well suited to a production situation.
• The basic philosophy is that the production procedure (formulation and process)
is allowed to evolve to the optimum by careful planning and constant repetition.
• The process is run in a way such that it both produces a product that meets all
specifications and (at the same time) generates information on product
improvement.
9
10. THE SIMPLEX METHOD
• The simplex approach to the optimum is also an experimental method and has
been applied more widely to pharmaceutical systems.
• A simplex is a geometric figure that has one more point than the number of
factors. So, for two factors or independent variables, the simplex is represented
by a triangle. Once the shape of a simplex has been determined, the method
can employ a simplex of fixed size or of variable sizes that are determined by
comparing the magnitudes of the responses after each successive calculation.
•The initial simplex is represented by the lowest triangle; the vertices represent
the spectrophotometric response. The strategy is to move toward a better
response by moving away from the worst response.
10
11. the worst response is
0.25,
conditions are selected at
the vortex, 0.6, and,
indeed,
improvement is obtained.
One can follow the
experimental path to the
optimum, 0.721.
Figure 5 The simplex approach to optimization. Response is spectorphotometric reading at a
given wavelength .
11
12. THE LAGRANGIAN METHOD
The several steps in the Lagrangian method can be summarized as follows:
1. Determine objective function
2 .Determine constraints
3. Change inequality constraints to equality constraints.
4. Form the Lagrange function, F:
a. One Lagrange multiplier λ for each constraint
b. One slack variable q for each inequality constraint
5. Partially differentiate the Lagrange function for each variable and Set derivatives
equal to zero.
6. Solve the set of simultaneous equations.
7. Substitute the resulting values into the objective functions.
12
13. •This technique requires that the experimentation be completed before
optimization so that mathematical models can be generated.
•The experimental design here was full 3 square factorial, and , as shown
in Table- 1 nine formulations were prepared.
13
14. Polynomial models relating the response variables to the independent variable
were generated by a backward stepwise regression analysis program. The
analyses were performed on a polynomial of the form and the terms were retained
or eliminated according to standard stepwise regression techniques.
y = B0+B1X1+B2X2+B3X12+B4X22+B5X1X2
+B6X1X22+B7X12X2+B8X12X22
In Eq. (3), y represents any given response and Bi represents the regression
coefficient for the various terms containing levels of the independent variable. One
equation is generated for each response or dependent variable.
14
15. EXAMPLE FOR THE LAGRANGIAN METHOD
The active ingredient, phenyl-propanolamine HCl, was kept at a constant level,
and the levels of disintegrant (corn starch) and lubricant (stearic acid) were
selected as the independent variables, X1 and X2. The dependent variables
include tablet hardness, friability, volume, in vitro release rate, and urinary
excretion rate inhuman subject.
A graphic technique may be obtained from the polynomial equations, as follows:
15
16. Figure 6. Contour plots for the Lagrangian method:
(a) tablet hardness;
16
17. Figure 6. Contour plots for the Lagrangian method:
(b) dissolution (t50%)
17
18. • If the requirements on the final tablet are that hardness be 8-10 kg and t50%
be 20-33 min, the feasible solution space is indicated in Fig. 6c.
•This has been obtained by superimposing Fig. 6a and b, and several
different combinations of X1 and X2 will suffice.
Figure 6. Contour plots for the Lagrangian method:
c) feasible solution space indicated by crosshatched area
18
19. oA technique called sensitivity analysis can provide information so that the
formulator can further trade off one property for another. For sensitivity analysis
the formulator solves the constrained optimization problem for systematic changes
in the secondary objectives. For example, the foregoing problem restricted tablet
friability, y3, to a maximum of 2.72%.
Figure 7 illustrates the in vitro release profile as this constraint is tightened or
relaxed and demonstrates that substantial improvement in the t50% can be obtained
up to about 1-2%.
19
20. Figure 7 illustrates the in
vitro release profile as this
constraint is tightened or
relaxed and demonstrates
that substantial improvement
in the t50% can be obtained
up to about 1-2%.
20
21. The plots of the independent
variables, X1 and X2, can be
obtained as shown in Fig.8.
Thus the formulator is
provided with the solution
(the formulation) as he
changed the friability
restriction.
Figure 8. Optimizing values of stearic acid and strach as a function of
restrictions on tablet friability: (A) percent starch; (B) percent stearic acid
21
22. Suspension design to illustrate
the efficient and effective
procedures that might be
applied. Representation of
such analysis and the
available solution space is
shown for the suspension in
Figs. 9 and 10.
Figure 9. Response surface concept and results of
the second case study
22
23. Figure 10. Secondary properties of various
suspensions yielding zero dose variation.
23
24. THE SEARCH METHOD
Although the Lagrangian method was able to handle several responses or
dependent variable, it was generally limited to two independent variables.
A search method of optimization was also applied to a pharmaceutical
system. It takes five independent variables into account and is computer-
assisted. It was proposed that the procedure described could be set up
such that persons unfamiliar with the mathematics of optimization and
with no previous computer experience could carry out an optimization
study.
25. THE SEARCH METHODS
1. Select a system
2. Select variables:
a. Independent
b. Dependent
3. Perform experimens and test product.
4. Submit data for statistical and regression analysis
5. Set specifications for feasibility program
6. Select constraints for grid search
7. Evaluate grid search printout
8. Request and evaluate:.
a. “Partial derivative” plots, single or composite
b. Contour plots
26. o The system selected here was also a tablet formulation .
The five independent variables or formulation factors selected for
this study are shown in Table 2.
28. • The
experimental design used was a modified factorial and is
shown in Table4.
• The fact that there are five independent variable dictates that a
total of 27 experiments or formulations be prepared. This design is
known as a five-factor, orthogonal, central, composite, second-order
design . The firs 16 formulations represent a half-factorial design
for five factors at two levels, resulting in ½ X 25 =16 trials. The two
levels are represented by +1 and -1, analogous to the high and low
values in any two level factorial design. For the remaining trials,
three additional levels were selected: zero represents a base level
midway between the aforementioned levels, and the levels noted as
1.547 represent extreme (or axial) values.
29.
30. • The translation of the statistical design into physical units is shown in
Table 5.
• Again the formulations were prepared and the responses measured. The
data were subject to statistical analysis, followed by multiple regression
analysis. This is an important step. One is not looking for the best of the
27 formulations, but the “global best.”
31. The type of predictor equation usd with this type of design is a second-
order polynomial of the following form:
Y = a 0+a1X1+…..+a5X5+a11X12+…+a55X52
+a12X1X2+a13X1X3+…+a45X4X5
Where Y is the level of a given response, the regression coefficients for
second-order polynomial, and X1 the level of the independent variable.
The full equation has 21 terms, and one such equation is generated for
each response variable
32. For the optimization itself, two major steps were used:
1. The feasibility search
2. The grid search.
The feasibility program is used to locate a set of response constraints that
are just at the limit of possibility.
. For example, the constraints in Table 6 were fed into the computer and
were relaxed one at a time until a solution was found.
33. This program is designed so that it stops after the first possibility, it is not a full
search.
The formulation obtained may be one of many possibilities satisfying the constraints.
34. •The grid search or exhaustive grid search, is essentially a brute
force method in which the experimental range is divided into a grid
of specific size and methodically searched.
•From an input of the desired criteria, the program prints out all
points (formulations) that satisfy the constraints.
• Graphic approaches are also available and graphic output is provided
by a plotter from computer tapes.
35. •The output includes plots of a given responses as a function of a single variable (fig.11).
The abscissa for both types is produced in experimental units, rather
than physical units, so that it extends from -1.547 to + 1.547.
36. The output includes plots of a given responses as a function of all five variable
(Fig 12).
37. Contour plots (Fig.13) are also generated in the same manner. The specific
response is noted on the graph, and again, the three fixed variables must be
held at some desired level. For the contour plots shown, both axes are in
experimental unit (eu) .
37
38. CANONICAL ANALYSIS
Canonical analysis, or canonical reduction, is a technique used to reduce a
second-order regression equation, to an equation consisting of a constant
and squared terms, as follows:
Y = Y0+λ1W12+λ2W22+…….
38
39. . In canonical analysis or canonical
reduction, second-order regression
equations are reduced to a simpler
form by a rigid rotation and translation
of the response surface axes in
multidimensional space, as shown in
Fig.14 for a two dimension system.
39
40. VI. OTHER APPLICATIONS
Formulation and Processing
Clinical Chemistry
Medicinal Chemistry
High Performance Liquid Chromatographic Analysis
Formulation of Culture Medium in Virological Studies.
Study of Pharmacokinetic Parameters.
40
41. . The graphs in Fig.15 show that for the drug hydrochlorothiazide, the
time of the plasma peak and the absorption rate constant could, indeed,
be controlled by the formulation and processing variables involved.
41
42. IX. REFERENCES
1. Websters Marriam Dictionary, G & C Marriam.
2. L. Cooper and N. Steinberg, Introduction to Methods of Optimization, W.B. Sunder.
3. O.L.Davis, The Design and Analysis of the Indusrial Experimentation, Macmillan.
4. Gilbert S. Banker, Modern Pharmaceutics, Marcel Dekker Inc.
5. Google search engine, WWW.Google.co.in
6. http://en.wikipedia.org/wiki/Optimization_(mathematics)
7. http://mat.gsia.cmu.edu/classes/QUANT/NOTES/chap4/node6.html
8. http://www.socialresearchmethods.net/kb/desexper.php
9. P .K. Shiromani and J. Clair, Drug Dev Ind Pharm., 26 (3), 357 (2000).
42